Radeon VII, what do you guys think ?

I hadn't considered the 1070 cards however they appear to share the same memory subsystem 256bit bus and 8GB GDDR5.

The only significant difference at the 1070 level seems to be the GPU in the Ti version has more CUDA cores 2431 vs 1920, higher clock about 7% over base clock, more texture units 152 vs 120 and TDP 180W vs 150W.

I'd be surprised if the 1070 Ti version didn't fair better with the same memory clock and amount of VRAM as reflected here in this link.

There is of course a difficulty inferring the effect that the amount of VRAM has on performance given the significant differences associated with the different GPU architectures 1080/2080/etc.

For instance if you reconfigured a 1070x to 6GB RAM does that infer 192bit bus and what impact would that have in it's own right?

Although from observation of DCS in game performance it doesn't seem to different between using say 7GB VRAM usage in SP and 10.something VRAM usage in MP, also from observation I could probably say the same for RAM usage as well in context of testing in the same version.

DCS has been all over the place in regards to how much system RAM and VRAM it uses with the various release iterations.

I'd be surprised if the 1070 Ti version didn't fair better with the same memory clock and amount of VRAM as reflected here in this link.

There is no doubt a 1070 Ti will do better than a 1070. The question is, how much. My speculation is that it will be more significant than most other games, because of the significantly higher texel fill rate (which doesnt do much for most AAA games), and a 1070 Ti should be pretty close to a 1080.

Admittedly, the result will still be difficult to interpret, as the Ti not only adds TMUs and ROPs, but also more shaders. My theory is that shader performance is not a major concern for DCS, but thats kinda hard to prove when all cards with higher texel fillrate also have higher shader count or performance..

Oh well. I guess that means it doesnt really matter if you shop for the wrong metric, if the right metric is included anyway ;).

I do think you have a point but unfortunately the "extras" are all part of "the package", I guess my point on VRAM capacity is that more VRAM usually means wider a bus which brings higher memory throughput. Besides being able to load more data into VRAM should at least help DCS overall.

Interestingly the Radeon VII whilst not as good a performer in general with AAA titles even against the 2080 might and I do stress might be advantageous in DCS VR with 3840 vs 2944 shader units.

@FragBum and/or all others

What do you think of the Radeon VII as an upgrade from a 1070 with respect to 4K resolution (3840 x 2160) performance?

To be honest I don't have any hard data for the Radeon VII for VR I think it has potential but I have no experience with it.

Other people with a 1070 seem to to get okay VR performance but then the whole DCS VR performance depends on CPU clock faster is better, GPU more grunt is better and RAM speed again faster is better.

It would be great for Radeon VII VR users to chime in with their experience, I just can't currently afford to buy one to test it out at this time but I would love to.

From reading various reviews the Radion VII is just piped at the post by 2080's and 1080Ti's but it also has some advantage in different scenarios. My main interest is whether the faster RAM and larger RAM capacity gives the Radeon VII an advantage in VR,

It also has more shader units compared to the 2080 which again should benefit DCS VR.

Sorry I just can't say yes or no at this time.

To be honest I don't have any hard data for the Radeon VII for VR I think it has potential but I have no experience with it.

 

Other people with a 1070 seem to to get okay VR performance but then the whole DCS VR performance depends on CPU clock faster is better, GPU more grunt is better and RAM speed again faster is better.

 

It would be great for Radeon VII VR users to chime in with their experience, I just can't currently afford to buy one to test it out at this time but I would love to.

 

From reading various reviews the Radion VII is just piped at the post by 2080's and 1080Ti's but it also has some advantage in different scenarios. My main interest is whether the faster RAM and larger RAM capacity gives the Radeon VII an advantage in VR,

 

It also has more shader units compared to the 2080 which again should benefit DCS VR.

 

Sorry I just can't say yes or no at this time.

Thanks for the feedback. I'm wondering if jumping from 8GB to 16GB VRAM will make for a more fluid 4K experience at UHD resolutions. Tried VR with a Rift and did not like the heat or scuba mask effect.

Nothing wrong with scuba diving in the game room,.. but definitely on those barmy tropical days. :P:D

Yes 4K again I haven't seen any hard data for the Radeon VII in DCS @ 4K I think the same sentiments apply the Radeon VII is a good card but at 4K it still generally gets piped at the post by the RTX 2080. Having said that I see in some reviews it does keep up with the RTX 2080 and in some cases allows more eye-candy, maybe because of faster RAM?

Problem is would DCS being so CPU dependent allow the Radeon VII to be fully utilised? From what else I've gleaned on the interwebs the Radeon VII would have advantages for maybe video editing which might be acceptable at it's slightly lower cost compared to RTX 2080.

Maybe wait for AMD Navi or even new nVidia offerings. Most people using the Radeon VII here seem happy enough with them though.

Maybe people with Radeon VII GPUs running DCS at 4K could chime in with some first hand wisdom. Although I suspect either way more CPU grunt might be needed as well. :) :thumbup:

Backy, are you sure you are GPU limited?

Like you, I was looking to upgrade my 1070, but Ive done a few tests in DCS and found that depending on the test, Im either overwhelmingly or entirely CPU limited. In my tests cpu speed and ram speed is pretty much all that matters, and even underclocking my GPU has no impact on framerates. I would probably not see much difference if I used a 1060.

Of course, I run at 1080p and 4K is different, but if you cant achieve good framerates at 1080p, upgrading your GPU isnt going to give good framerates at 4K.

If you havent seen it already:

https://forums.eagle.ru/showthread.php?t=240207

https://forums.eagle.ru/showthread.php?t=240131

Especially the first thread proves how much we are bottleneck by the CPU and ram. Someone reran the test at 4K with his GTX 1080, and pretty much achieved identical results compared to 1080p

instead of upgrading my GPU, I ordered DDR4-3466, and I will wait to see if Ryzen3 brings anything to the table.

But the most significant upgrade for all of us could well be the Vulkan API.

Backy, are you sure you are GPU limited?

 

Like you, I was looking to upgrade my 1070, but Ive done a few tests in DCS and found that depending on the test, Im either overwhelmingly or entirely CPU limited. In my tests cpu speed and ram speed is pretty much all that matters, and even underclocking my GPU has no impact on framerates. I would probably not see much difference if I used a 1060.

 

Of course, I run at 1080p and 4K is different, but if you cant achieve good framerates at 1080p, upgrading your GPU isnt going to give good framerates at 4K.

 

If you havent seen it already:

https://forums.eagle.ru/showthread.php?t=240207

https://forums.eagle.ru/showthread.php?t=240131

 

Especially the first thread proves how much we are bottleneck by the CPU and ram. Someone reran the test at 4K with his GTX 1080, and pretty much achieved identical results compared to 1080p

 

instead of upgrading my GPU, I ordered DDR4-3466, and I will wait to see if Ryzen3 brings anything to the table.

 

But the most significant upgrade for all of us could well be the Vulkan API.

Thanks for the links to the test threads Vertigo. I will check it out. I suspect my i7-6700 could be my weak link since I can't overclock it.

I cant really overclock my CPU either, but instead of suspecting, do as I did: underclock. Both CPU and GPU (one at the time). If framerates are not affected by the underclock of the component you are underclocking, the bottleneck is elsewhere. If framerates drop linearly with clockspeeds, then you are entirely bottleneck by that component

Radeon VII + Ultrawide

Hi Bit,

I finally upgraded both GPU (from 1060 6Gb to a Sapphire Radeon VII) and monitor (from 27'' HD 60Hz to a 34'', 144Hz, 3440x1440 beast, the LG 34GK950F). You may remember helping me selecting components for my rig last year.

I am still playing around with all the GPU/Monitor settings, but the improvement in fps and smoothness out of the box is very, very large.

First things first: with all settings high, 2x MSAA, no anis. filtering, FreeSync active, etc.. I will take a screenshot when back home) coupled with the GPU stock boost freq. (1800MHz) and a little undervolting (for now at 1050mV down from 1123mV, still stable, will push further in the next days) I can firmly stay in the 60-80fps range, with GPU core temp below 60deg and junction temp always below 100deg. I was seeing 112deg with stock settings....

No stuttering whatsoever, some drops to 45-50fps when taxiing. I've narrowed this down to grass/clutter density (could verify a linear correlation), and I've now set it to 750 to avoid dropping further.

When the GPU junction temp raises, so does the fan noise. I understand that the fan RPMs are coupled to the junction temp now, not the core. I do not mind it, since I use earphones in DCS, but I could still hear the blowing ! It is not a whistle, but a low-pitch noise similar to the one of a room ventilator spinning at high speed. It does not bother me, but this is subjective. There are already some mods around for improving the dissipation with the stock fans, but I know I will water-cool it soon, just to be able to play with OCs settings with safer margins.

If you guys want some more details/tests, just let me know and I will run them for you. My specs are as follows

i7 8700k (water-cooled, 4.7GHz, not OC)

32Gb RAM @3200MHz

Sapphire Radeon VII (could try different max Freq./Voltage)

Ultrawide 3440x1440 @120Hz (could go at 144Hz, I do not see the benefit in DCS though)

Modules: A10, F18, P51, Spitfire; all maps.

nice !

If you have the tomcat module, I would be interested in seeing your results in this test:

https://forums.eagle.ru/showthread.php?t=240207

Although the test doesnt seem to be GPU bound for almost anyone, instead its CPU single thread and CPU IO bound, Im under the impressing AMD GPUs do better, possibly because their drivers incur a lower CPU overhead.

If you dont have the tomcat, then maybe you can post your result for this test here:

https://forums.eagle.ru/showthread.php?t=240131

If you have the tomcat module, I would be interested in seeing your results in this test:

https://forums.eagle.ru/showthread.php?t=240207

 

Although the test doesnt seem to be GPU bound for almost anyone, instead its CPU single thread and CPU IO bound, Im under the impressing AMD GPUs do better, possibly because their drivers incur a lower CPU overhead.

 

If you dont have the tomcat, then maybe you can post your result for this test here:

https://forums.eagle.ru/showthread.php?t=240131

Sure, I'll run the second one, as I don't have the Tomcat. I'll post the results there and let you know here. EDIT: Done, reported there. Good overall results, GPU maximized almost all the time (drops in external views), temperatures strangely low, CPU never maximized (yes, I am overlaying utilization of all cores, not one is ever above 90%). This is baffling, when compared to what others are reporting.

I've run a few more tests on native resolution with own settings, I'll try to assemble the results in a separate thread. Spoiler:

23-05-2019, 00:59:35 DCS.exe benchmark completed, 15289 frames rendered in 212.656 s   
                    Average framerate  :   71.8 FPS
                    Minimum framerate  :   23.9 FPS
                    Maximum framerate  :  143.6 FPS
                    1% low framerate   :   14.7 FPS
                    0.1% low framerate :   11.6 FPS

(Starting from HIGH preset, applying these modifications: Vsync on, clutter/grass down to 0, No Anisotropic filt., Shadows Flat, No MSAA, No SSAA, Visibility range to extreme, firt+flare ON)

Radeon settings: VSync enhanced; MSAA and Anisotropic filtering maximized at 8xEQ and 16x, respectively.

Resolution at 3440x1440.

Pretty, pretty nice visually, and mostly >60 fps on ground (granted, no grass/clutter).

Radeon VII or GTX 2080???

Hi Bit,

I finally upgraded both GPU (from 1060 6Gb to a Sapphire Radeon VII) and monitor (from 27'' HD 60Hz to a 34'', 144Hz, 3440x1440 beast, the LG 34GK950F). You may remember helping me selecting components for my rig last year.

I am still playing around with all the GPU/Monitor settings, but the improvement in fps and smoothness out of the box is very, very large.

First things first: with all settings high, 2x MSAA, no anis. filtering, FreeSync active, etc.. I will take a screenshot when back home) coupled with the GPU stock boost freq. (1800MHz) and a little undervolting (for now at 1050mV down from 1123mV, still stable, will push further in the next days) I can firmly stay in the 60-80fps range, with GPU core temp below 60deg and junction temp always below 100deg. I was seeing 112deg with stock settings....

No stuttering whatsoever, some drops to 45-50fps when taxiing. I've narrowed this down to grass/clutter density (could verify a linear correlation), and I've now set it to 750 to avoid dropping further.

When the GPU junction temp raises, so does the fan noise. I understand that the fan RPMs are coupled to the junction temp now, not the core. I do not mind it, since I use earphones in DCS, but I could still hear the blowing ! It is not a whistle, but a low-pitch noise similar to the one of a room ventilator spinning at high speed. It does not bother me, but this is subjective. There are already some mods around for improving the dissipation with the stock fans, but I know I will water-cool it soon, just to be able to play with OCs settings with safer margins.

 

If you guys want some more details/tests, just let me know and I will run them for you. My specs are as follows

i7 8700k (water-cooled, 4.7GHz, not OC)

32Gb RAM @3200MHz

Sapphire Radeon VII (could try different max Freq./Voltage)

Ultrawide 3440x1440 @120Hz (could go at 144Hz, I do not see the benefit in DCS though)

Modules: A10, F18, P51, Spitfire; all maps.

Hey I'm looking to upgrade my system and I'm really thinking the 16GB of VRAM should be really helpful long term but the GTX 2080 seems to be more stable and higher frame rates.... what would you suggest I get?

Hey I'm looking to upgrade my system and I'm really thinking the 16GB of VRAM should be really helpful long term but the GTX 2080 seems to be more stable and higher frame rates.... what would you suggest I get?

For what I understand, they seem to be pretty much on par when it comes to DCS. But to my knowledge nobody has done any direct comparison, so I may be wrong. As for stability, I did not have any GPU crash-related problem since I got it, except of course when I was fiddling with voltage/frequency. I am now settled at 1000mV/1800MHz, with some great performance after the latest updates in Beta: I rarely go below 70fps and most of the time I'm in the high 80's with almost all maxed out, on an ultrawide 3440x1440.

Probably a 2080 would get similar results, with marginal plus or minuses. For me, the choice was dictated by a combination of low price for the VII and a genuine dislike for the counterpart offering.

So, there, I can't really help you choosing, just offering my positive review of the VII, it has been nothing but good for me so far. Remember: the fans are loud .. when I play the Witcher (which seems to be a lot more taxing to the GPU than DCS, btw) there is a howling wolf screaming from inside my rig. Luckily, it adds to the atmosphere of the game